This is the practical scaffold we used informally in IAP 2023: keep the book chapters in order, but pause each meeting to run code or a proof sketch so abstractions stay moored to systems.
- Ground rules. OCW lists linear algebra maturity (18.06) and MATLAB familiarity. If you live in Python, still skim MATLAB for D4M idioms first, then port once the algebra is automatic.
- Session 0 — Install and sanity-check. Download D4M from d4m.mit.edu, clone a demo such as the baseball walkthrough, and read the ICASSP 2012 D4M paper once for vocabulary.
- Weekly rhythm. For each book chapter: (a) read the chapter; (b) watch the matching arc from RES.LL-005 (Fall 2012 video plus IAP 2020 supplements); (c) implement at least one example or exercise; (d) write a short note translating the idea into a system you care about—warehouse, stream, graph, simulation mesh.
- Bridge to modern graphs. After Part II (rigorous associative-array definitions), read the GraphBLAS community materials and one GraphChallenge champion write-up. The goal is to recognize the same multiplication pattern on different substrates.
- Capstone. Pick one tractable GraphChallenge task or sparse benchmark and build a toy implementation. Aim for performance literacy, not leaderboard placement.
OCW's lecture notes page bundles IAP 2020 PDF supplements on AI/ML and cyber data architecture—good bridge readings when the textbook feels pure-math heavy.